Full Field Chemical Imaging of Buried Native Sub-Oxide Layers on Doped Silicon Patterns
Total Page:16
File Type:pdf, Size:1020Kb
Surface Science 604 (2010) 1628–1636 - doi : 10.1016/j.susc.2010.06.006 Published in Elsevier Surface Science Full field chemical imaging of buried native sub-oxide layers on doped silicon patterns a b, a a c F. de la Peña , N. Barrett , L.F. Zagonel , M. Walls , O. Renault a Univ Paris-Sud, Laboratoire de Physique des Solides, CNRS/UMR 8502, Orsay, F-91405, France b CEA, IRAMIS, SPCSI, LENSIS, F-91191 Gif-sur-Yvette, France c CEA-LETI, MINATEC, 17 rue des Martyrs, 38054 Grenoble Cedex 9, France DOI: 10.1016/j.susc.2010.06.006 a b s t r a c t Keywords: Fully energy-filtered X-ray photoelectron emission microscopy is used to analyze the Semiconductor–insulator interfaces spatial distribution of the silicon sub-oxide structure at the SiO2/Si interface as a Semiconductor–semiconductor thin film function of underlying doping pattern. Using a spectroscopic pixel-by-pixel curve fitting structures analysis, we obtain the sub-oxide binding energy and intensity distributions over the full Silicon oxides field of view. Binding energy maps for each oxidation state are obtained with a spatial Photoelectron emission resolution of 120 nm. Within the framework of a five-layer model, the experimental Synchrotron radiation photoelectron data are used to obtain quantitative maps of the sub-oxide layer thickness and also their spectroscopy spatial distribution over the p–n junctions. Variations in the sub-oxide thicknesses are Photoelectron emission microscopy (XPEEM) found to be linked to the level and type of doping. The procedure, which takes into account instrumental artefacts, enables the quantitative analysis of the full 3D dataset. 1. Introduction presence of interface dipoles [2,3]. Finally, the width of the photoelectron spectrum may be measured and thus the work function deduced. The original In silicon based micro- and nanotechnologies, the control of the quality reference for synchrotron radiation-based PES studies of this system is the and thickness of the SiO2/Si interface is of prime importance. An interface of work of the Himpsel group on Si(100) and Si(111) [4]. Other experimental several atomic layers, with a high defect concentration can result in fixed studies have been carried out by [5,6]. Several models of the interface have charges and contribute significantly to variations in electrical properties. One been proposed in the light of theoretical calculations [7–10]. However, of the limits on CMOS downscaling is the quality of the interface. The typical X-ray spot sizes are of the order of a fraction of a millimetre, therefore chemistry of the SiO2/Si interface has already been studied in the laboratory virtually no spatial information is available. Thus, PES is an area-averaged using both angle-resolved X-ray photoelectron spectroscopy (XPS) with Al technique and hence cannot address gate oxide dimensions compatible with K or Mg K X-ray sources and synchrotron radiation-based photoelectron real circuits. Downscaling in microelectronics makes the availability of spectroscopy (PES). The use of XPS to obtain accurate values for the reliable spatially resolved quantification of the interface chemical states a thickness of the SiO2 layer and the sub-oxide interfaces has attained pressing requirement. extremely high standards in order to define a universal method independent of Energy resolved X-ray photoelectron emission microscopy (XPEEM) specific laboratory conditions [1]. The precision now possible in measuring combines the required spatial and chemical state resolution. In particular, by oxide layer thickness is smaller than typical variations in silicon oxygen bond using soft X-ray excitation combined with a full energy-filtered analysis, lengths. Photoelectron spectroscopy is also non-destructive, allowing elemental, chemical and electronic structure sensitivities of classical PES with complementary analyses such as Secondary ion mass spectrometry (SIMS) to spatial resolutions on the 100 nm scale are now readily accessible with be performed a posteriori. Synchrotron radiation is particularly useful for a current PEEM instruments. Image series as a function of photoelectron kinetic detailed analysis of the SiO2/Si interface since the depth probed can be energy Ek are acquired step by step over the energy window of interest. Each changed by tuning the photon energy. Valence band PES gives information image gives the intensity as a function of position within the microscope field on the density of states and the valence band offsets, in particular on the of view (FoV). A full image series is a 3D dataset, called a Spectrum Image (SI), allowing extraction of photoemission spectra within the FoV from any area of interest (AOI) down to the pixel size. This technique was first applied in a study of silicon anodic oxidation [11], in which the effect Corresponding author. Tel.: +33 169083272; fax: +33 169088446. E-mail address: [email protected] (N. Barrett). F. de la Peña et al. / Surface Science 604 (2010) 1628–1636 doi : 10.1016/j.susc.2010.06.006 1629 2. Experiment The samples, made in the Laboratoire d'Electronique et de Technologie de l'Information (LETI) at MINATEC, CEA-Grenoble, consisted of two- dimensional doped patterns implanted into silicon in the form of lines with variable spacing and widths from 100m to 0.1m [12]. Before implantation, the silicon wafers were clean and had any surface oxide removed by standard chemical procedures. Each set of lines was identified by figures with the same doping level, giving suitable shapes for optimization of the electron optics of the PEEM instrument. Two samples were studied, Fig. 1. Phosphor and boron ions were used for the n- and p-type doping. The first sample, designated N+/P−, had N+ (representing very high n-type doping) doped zones (1020 cm−3) implanted into a P− substrate (1016 cm−3). The second, designated P+/N, had P+ (representing very high p-type doping) patterns (1020 cm−3) implanted into an N type (1017 cm−3) substrate; the latter was itself 16 −3 deposited on a Si P− wafer (10 cm ). A native oxide layer was present on both samples. Fig. 1 also illustrates schematically the expected band line-ups in such samples. The two samples are not symmetric in their preparation since in both cases the starting substrate is P−. Hence, the P+/N sample underwent a supplementary ion implantation to obtain the required doping patterns. The + − doping levels have been measured using secondary ion mass spectrometry Fig. 1. Optical micrograph of doped pattering schematic cross-section of samples N /P and + (SIMS); the results are given in Table 1. P /N. The oxide/substrate and p–n band line-ups are also shown. The spectromicroscopy experiments used a NanoESCA XPEEM system (Omicron Nanotechnology), more fully described elsewhere [16,18]. The of electrostatic charge on the Si 2p emission due to the anodic oxide was apparatus was installed on the CIPO beamline of the ELETTRA synchrotron demonstrated. Using this method, a preliminary analysis of the present system (Trieste, Italy). The incident photon energy was 127 eV, representing the best revealed doping-dependent variations in the overall oxide layer thickness compromise between the beamline undulator response and surface sensitivity. [12]. However, depending on the size of the AOIs up to 99% of the At this energy, the inelastic mean free path (IMFP) of the Si 2p electrons is information in the dataset is not used. low, thus the substrate signal is expected to reflect bent bands at the The most information is available from pixel-by-pixel curve fitting to substrate/oxide interface rather than the flat band situation. Given the depth spectra over the full field. This has already been used in XPEEM analysis to sensitivity, photoelectron diffraction contributions from the substrate are also obtain, for example, concentration maps of InAs/GaAs quantum dots and negligible. The photoelectrons are collected within a 6.1° cone around the rings [13,14]. Ga 3d and In 4d core levels, separated by 2 eV were resolved surface normal. The PEEM contrast aperture diameter was 150 μm, a trade- allowing elemental mapping of the surface. An alternative full field approach off between lateral resolution and PEEM column transmission. The double was developed by Ratto et al. [15], however, it relies on the existence of an internal reference in the sample with known composition, which is not always the case. Furthermore, the information extracted is averaged over the photoelectron escape depth. Such analysis is often, but not always, limited to elemental mapping based on well separated core levels. To accurately map the surface and sub-surface spatial variations in the chemical states and binding energies, careful handling of the dataset and flexible numerical analysis tools are required. One must correct for the microscope transmission function (or illumination) at constant kinetic energy in the FoV, the variations in the photon flux and the non-isochromaticity over the FoV [16,17]. When one does this, the inevitably better statistics gives new quantitative information on the spatial distribution of the chemical states and on sub- surface interfaces. As a test case, we have studied the interface structure of the native oxide as a function of the Si(100) substrate doping. The area- averaged Si 2p core-level studies have already shown that the interface structure is strongly dependent on the particular oxide growth conditions [4,5]. We obtain maps of the SiO2 and the sub-oxide distributions. We show that the SiO2/Si interface is not the same over p- and n-type doped patterns, and develop a model to quantify the spatial extent of the different interface sub-oxides between SiO2 and Si. Table 1 Doping levels as measured by secondary ion mass spectrometry. 4+ Fig. 2. (a) Non-isochromaticity measurement by a parabolic fit to the Si peak energy as a Sample n-type (at/cm³) p-type (at/cm³) function of the vertical (energy dispersive plane) pixel position in a region far from p–n + − 4+ 20 15 junctions for sample N /P .